MIT Creates Twitterbot 'Deep Drumpf' That Tweets Better Than Trump

By Ana Verayo, | March 06, 2016

MIT created a Twitterbot called DeepDrumpf, tweeting like Trump himself.

MIT created a Twitterbot called DeepDrumpf, tweeting like Trump himself.

A Twitterbot was developed in the likeness of celebrity turned U.S. Republican presidential candidate Donald Trump that tweets statements that sound like Trump himself, courtesy of a researcher from the Massachusetts Institute of Technology (MIT).

Like Us on Facebook

This bot is known as "Deep Drumpf" which is based on an artificial intelligence algorithm that have been fed with only a few hours of transcripts of Trump's speeches and debate performances, according to Bradley Hayes of MIT's Computer Science and Artificial Intelligence Lab (CSAIL).

The bot's name "Deep Drumpf" is originally from comedian John Oliver and host of "Last Week Tonight", who also had Trump as a guest on the show, when Trump eventually revealed that his family's original name is apparently Drumpf. 

The bot decides on what to tweet at random, where it tweets one letter at a time. For example, if the bot begins with the letter "M", the next letter that would be typed is probably "A" and then will be followed by a "K" to complete the campaign slogan of Trump, "Make America Great Again". The same process happens again for the next sentence until it reaches the 140 character limit per tweet.

Hayes utilized a method from "deep learning", which is a system that is used in the artificial intelligence field for their neural networks, that enables computers to recognize patterns on their own capacity.

The idea for this Trump Twitterbot comes from an existing training model simulating Shakespeare and also a recent report that analyzed the linguistic patterns of presidential candidates, that revealed how Trump speaks sentences at a fourth grade level.

Hayes reveals that Trump's language appears to be more simple in form, so figuring out that as a modelling problem, Trump became the most manageable candidate for further study.

Most of tweets created by Deep Drumpf do not always make complete sense but they are somewhat, partially coherent, much like Trump himself.

Hayes further explains that this algorithm learns from an underlying structure derived from all the data it can collect, where it comes up with diiferent combinations of that data, reflecting this specific structure that it was trained to do. 

Deep Drumpf even directly interfaced with the real Twitter account of Trump, where in these cases, Hayes provides an algorithm language derived from the tweet of the real Trump which will then be primed to give out a response that is more or less relevant in context.

©2024 Telegiz All rights reserved. Do not reproduce without permission
Real Time Analytics